Find where your LLM API bill leaks—without uploading your logs

Описание и применение

linklbee‑scan – это локальный инструмент на Python, который анализирует журналы использования LLM и генерирует HTML‑отчёт о расходах, выявляя повторяющиеся запросы, низкую эффективность кэша и обрезанные ответы, что позволяет оптимизировать затраты без передачи данных сторонним сервисам.

Задача

Determine which parts of an LLM API bill are driving costs and identify opportunities for savings without exposing sensitive logs.

Результат

linklbee‑scan produces an HTML report that identifies cost drivers such as repeated calls, low cache hit rates, and truncated responses, allowing teams to target specific experiments for savings.

Рабочий процесс

• Install linklbee‑scan via pip
• Run the built‑in demo to view report format
• Provide your own log file or folder to linklbee‑scan
• Optionally use --peek to verify field mapping
• Review the generated report for cost patterns and findings
• Turn findings into experiments (e.g., reuse identical calls, adjust prompt length)
• Compare cost, quality, and latency before and after changes

Доказательства

We built Linklbee Scan, an open‑source Python tool that reads existing usage logs on your machine and generates an HTML report to help answer those questions.

Источник

No API key, no model calls, and no log uploads. It uses the Python standard library and runs on Python 3.9+.

Источник

The scanner reads JSONL, JSON arrays, CSV/TSV, .log, and .gz files. It includes mappings for common provider responses and exports from tools such as Langfuse, LangSmith, Helicone, and LiteLLM.

Источник